The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: Potential risk of indirect prompt injection due to processing untrusted web content.
Ingestion points: The skill fetches external data from social media, forums, and blogs using Exa search tools (web_search and deep_researcher) as specified in SKILL.md and references/exa-queries.md.
Boundary markers: No specific delimiters or instructions are provided to isolate untrusted web search results from the system prompt or to prevent the agent from following adversarial instructions embedded in community posts.
Capability inventory: The skill contains no executable code or scripts, relying entirely on the agent's markdown processing, reporting capabilities, and the external Exa tool context.
Sanitization: Content from external sources is directly parsed for sentiment analysis and interpolated into template variables ({quote}) without sanitization or escaping filters in references/templates/pulse-report.md.

community-pulse